stanova_samples: Posterior samples for ANOVA-type models

In bayesstuff/stanova: Bayesian Models with Categorical Variables

Description

Posterior samples for ANOVA-type models

Usage

stanova_samples(object, ...)

## S3 method for class 'stanova'
stanova_samples(
  object,
  diff_intercept = TRUE,
  terms,
  return = c("array", "matrix", "data.frame", "tibble", "tidybayes"),
  dimension_chain = 3L,
  ...
)

Arguments

object	Fitted model. Currently only support for stanova objects.
...	currently ignored.
diff_intercept	logical. If TRUE (the default) samples for factor levels represent the difference from the intercept. If FALSE, marginal means of factor levels are returned.
terms	optional character vector denoting a subset of model terms for which difference from intercept be returned.
return	character string denoting format in which samples should be returned. Allowed values are c("array", "matrix", "data.frame", "tibble", "tidybayes"), possibly abbreviated. "tibble" requires package tibble and returns the same object as data.frame wrapped into as_tibble(). "tidybayes" also returns essentially the same object, but renames the columns to follow tidybayes conventions.
dimension_chain	Scalar integer. If return = "array", determines the dimension of the chain. The default 3 means chain is the third dimension of the returned array. Value should be between 1 and 3.

Examples

fit_warp <- stanova(breaks ~ wool * tension, data = warpbreaks,
                    prior = rstanarm::R2(0.5),
                    model_fun = "lm",
                    chains = 2, iter = 500)

arr_warp <- stanova_samples(fit_warp)
str(arr_warp)

mat_warp <- stanova_samples(fit_warp, return = "matrix")
str(mat_warp)

df_warp <- stanova_samples(fit_warp, return = "data.frame")
str(df_warp)

tail(df_warp$`wool:tension`)

if (requireNamespace("tibble")) {
  tbl_warp <- stanova_samples(fit_warp, return = "tibble")
  tbl_warp
}

if (require("tidybayes") && require("tidyverse")) {
  tidy_warp <- stanova_samples(fit_warp, return = "tidybayes")
  tidy_warp

  tidy_warp$tension %>%
    group_by(variable) %>%
    median_qi(value) %>%
    ggplot(aes(y = variable, x = value, xmin = .lower, xmax = .upper)) +
    geom_pointinterval()

  tidy_warp$tension %>%
    ggplot(aes(y = variable, x = value)) +
    stat_halfeye()

  bind_rows(tidy_warp) %>%
    ggplot(aes(y = variable, x = value)) +
    stat_halfeye() +
    facet_wrap(vars(term), scales = "free")

   bind_rows(tidy_warp) %>%
    ggplot(aes(y = variable, x = value)) +
    stat_interval() +
    facet_wrap(vars(term), scales = "free")

   ## marginal means instead of differences from mean:
   stanova_samples(fit_warp, return = "tidybayes", diff_intercept = FALSE) %>%
     bind_rows() %>%
     ggplot(aes(y = variable, x = value)) +
     stat_halfeye() +
     facet_wrap(vars(term), scales = "free")
}

bayesstuff/stanova documentation built on June 9, 2021, 6:18 p.m.